Struct CodeParser 

pub struct CodeParser { /* private fields */ }

A robust code parser system that analyzes source code and produces clean, accurate Abstract Syntax Trees (ASTs) optimized for LLM consumption.

Key capabilities:

• Consistent parsing approach using tree-sitter for reliable, accurate parsing
• Clean, well-documented API for LLM tool use
• Efficient error recovery for handling malformed code
• Structured AST output that LLMs can easily interpret
• Language detection with robust extension mapping
• Declarative query patterns for extracting meaningful code structures
• Efficient caching system for parsers and queries

Implementations

impl CodeParser

pub fn new() -> Result<Self>

Creates a new CodeParser instance with initialized language support and default configuration.

Returns

• Result<Self> - A new CodeParser instance or an error

pub fn with_config( cache_size_limit: Option<usize>, max_file_size: Option<usize>, max_files: Option<usize>, max_depth: Option<usize>, ) -> Result<Self>

Creates a new CodeParser instance with custom configuration.

Arguments

• cache_size_limit - Optional cache size limit in bytes (default: 50MB)
• max_file_size - Optional maximum file size to parse in bytes (default: 1MB)
• max_files - Optional maximum number of files to parse (default: 25)
• max_depth - Optional maximum recursion depth (default: 3)

Returns

• Result<Self> - A new CodeParser instance or an error

pub fn detect_language(&self, path: &Path) -> Option<String>

Determines the programming language from a file extension

Arguments

• path - Path to the file

Returns

• Option<String> - Language name if detected

pub fn parse_file(&mut self, path: &Path) -> Result<CodeAST>

Parses a single file using tree-sitter and generates an AST

Arguments

• path - Path to the file to parse

Returns

• Result<CodeAST> - The abstract syntax tree or an error

pub fn create_simplified_ast( &self, path: &Path, language: &str, source_code: &str, ) -> Result<CodeAST>

Create a simplified AST directly from the source code This is a fallback method when tree-sitter parsing doesn’t work

Arguments

• path - Path to the file
• language - Language of the source code
• source_code - Source code text

Returns

• Result<CodeAST> - Simplified AST or error

pub fn extract_search_terms(&self, query: &str) -> Vec<String>

Extract search terms from a query for grep search

Arguments

• query - User query string

Returns

• Vec<String> - Extracted search terms

pub fn parse_codebase( &mut self, root_dir: &Path, query: &str, ) -> Result<Vec<CodeAST>>

Parse an entire codebase and generate ASTs for selected files

Arguments

• root_dir - Root directory of the codebase
• query - User query to determine relevant files

Returns

• Result<Vec<CodeAST>> - List of ASTs for relevant files

pub fn generate_llm_friendly_ast( &mut self, root_dir: &Path, query: &str, ) -> Result<String>

Generate a structured AST optimized for LLM consumption

Arguments

• root_dir - Root directory of the codebase or path to a single file
• query - User query to determine relevant files

Returns

• Result<String> - Structured AST as a string

pub fn determine_relevant_files(&self, query: &str) -> Vec<String>

Determine which files to parse based on user query

Arguments

• query - User query string

Returns

• Vec<String> - List of glob patterns for relevant files